Neural network prediction of relativistic electrons at geosynchronous orbit during the storm recovery phase: effects of recurring substorms

During the recovery phase of geomagnetic storms, the flux of relativistic (> 2 MeV) electrons at geosynchronous orbits is enhanced. This enhancement reaches a level that can cause devastating damage to instruments on satellites. To predict these temporal variations, we have developed neural network models that predict the flux for the period 1-12h ahead. The electron-flux data obtained during storms, from the Space Environment Monitor on board a Geostationary Meteorological Satellite, were used to construct the model. Various combinations of the input parameters AL, SigmaAL, D-st and SigmaD(st) were tested (where Sigma denotes the summation from the time of the minimum D-st). It was found that the model, including SigmaAL as one of the input parameters, can provide some measure of relativistic electron-flux prediction at geosynchronous orbit during the recovery phase. We suo-est from this result that the relativistic electron-flux enhancement during the recovery phase is associated with recurring substorms after D-st minimum and their accumulation effect.